Imbalanced data is a type of data where there exists a difference in the ratio of classes. It occurs easily in real life of data analysis. In Data mining the functioning of learning algorithms caused by the imbalanced data. Most of the machine learning algorithms has a tendency to prejudice towards the class of majority in case of imbalanced data and hence those algorithms misjudge the minority class. Therefore, In this article we discuss a systematic way to address the imbalanced data classification problem by applying the rule based ensemble learning techniques like bagging, boosting, voting and stacking to build models, and then accelerates the performance of learning algorithms. In this research, we have preferred real data of chronic kidney disease which is collected from Appolo Hospitals, Tamil Nadu, India, to predict kidney disease of patients .The collected data is initially imbalanced. Firstly, the imbalanced data is balanced by applying SMOTE algorithm, which is an over sampling technique. Then applied various ensemble learning techniques to make better prediction. The incurred results showed that the model template chosen can minimize the problem of misclassification of imbalanced data efficaciously. But this model template cannot classify correctly when imbalanced rate of class increases i.e. in case of Big Data. For better result of imbalanced Big Data, new algorithmic plan of action has to be exploited which can be measured by using Hadoop framework and mapreduce programming model.
Feature Selection (FS) is an imperative issue in data mining and machine learning. It is an inevitable task to shorter the number of features presented in the initial data set for better classification result, minimized computation time, and reduced memory consumption. In this article, a novel framework using Correlation Coefficient (CCE) and Symmetrical Uncertainty (SU) for selecting the subset of feature is proposed. The selected features are congregated into finite number of clusters by grading their CCE and comparing the SU values. In each cluster, a feature with maximum SU value is retained while remaining features in the same cluster are ignored. The proposed framework was examined with Ten(10) real time benchmark data sets. Experimental outcomes show that the proposed method is outruns than majority of conventional feature selection methods(Information Gain, Chi-Square, Gain Ratio, ReliefF) in accuracy. This method is tested using Tree Based, Rule Based, Lazy, and Naive Bayes learners.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.